Research map / Image generation / Text-to-image diffusion
Lumina-Image 2.0: A Unified and Efficient Image Generative Framework
Qi Qin, Le Zhuo, Yi Xin, Ruoyi Du, Zhen Li, Bin Fu, Yiting Lu, Xinyue Li, Dongyang Liu, Xiangyang Zhu, Will Beddow, Erwann Millon and 8 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| scale training data from 20M to 200M samples | Lumina-Next · DPG | DPG overall: 75.66 -> 85.80 | Improved |
| add high-quality tuning stage after high-res stage (ablation) | Lumina-Image 2.0 · DPG | DPG overall: 85.7 -> 86.6 | Improved |
| add high-quality tuning stage after high-res stage (ablation) | Lumina-Image 2.0 · GenEval | GenEval overall: 0.67 -> 0.71 | Improved |
| train high-quality tuning from 5K to 11K steps (ablation) | Lumina-Image 2.0 · DPG | DPG overall: 87.2 -> 87.6 | Improved |
| train high-quality tuning from 5K to 11K steps (ablation) | Lumina-Image 2.0 · GenEval | GenEval overall: 0.73 -> 0.72 | Hurt |
| train high-res stage after low-res stage (ablation) | Lumina-Image 2.0 · DPG | DPG overall: 84.5 -> 85.7 | Improved |
| add CFG-Renorm and CFG-Trunc (ablation) | Lumina-Image 2.0 · 1024 resolution sampling | sampling time and quality: over a 20% acceleration | Improved |
| add Flow-DPM-Solver (ablation) | Lumina-Image 2.0 · 1024 resolution sampling | sampling time and sample quality | Mixed |
| add TeaCache (ablation) | Lumina-Image 2.0 · 1024 resolution sampling | speed and image quality | Mixed |
| use longer, more precise UniCap captions instead of short Florence captions (ablation) | Lumina-Image 2.0 · training loss | convergence speed | Improved |
| replace cross-attention with joint self-attention (Unified Next-DiT) plus UniCap and more data | Lumina-Next · DPG | DPG overall: 75.66 -> 87.20 | Improved |
Later papers that compared with this method
2 later accepted papers report Lumina-Image 2.0 as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Gen-Searcher: Reinforcing Agentic Search for Image Generation (NeurIPS 2026 · better on 1 of 1 shared result)
- T2I-Copilot: A Training-Free Multi-Agent Text-to-Image System for Enhanced Prompt Interpretation and Interactive Generation (ICCV 2025 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Sana-1.6B — SANA: Efficient High-Resolution Text-to-Image Synthesis with Linear Diffusion Transformers (ICLR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency (ICLR 2024 · similarity 0.88)
- CoBIT: A Contrastive Bi-directional Image-Text Generation Model (ICLR 2024 · similarity 0.88)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.88)
- Elucidating the design space of language models for image generation (ICML 2025 · similarity 0.87)
- OmniGen: Unified Image Generation (CVPR 2025 · similarity 0.87)
- MARS: Mixture of Auto-Regressive Models for Fine-grained Text-to-image Synthesis (AAAI 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
